We have developed a new approach for migration velocity analysis by ray-based reflection tomography, formulated according to more than a single dip direction. It is suggested to use a summation-free subsurface imaging system in the angle domain for generating multiparameter common image gathers through depth migration. Each gather associated with this system comprised dip-dependent opening-angle images contributed from the prestack data. Independent moveout information, coming from either a specular or nonspecular dip direction, was extracted inside these gathers to allow the entire scattered field to be involved in the velocity model optimization. By obeying the linear tomographic principle, a multidip tomography system was set to include imaging errors from specular and nonspecular directions. The updated velocity model was reconstructed by a least-squares inverse solution of the multidip tomographic equation system. Providing additional moveout from the migration’s dip, other than the specular one, was believed to be essential because the seismic data were misplaced in the image space while applying depth imaging by an erroneous velocity model. It might make the determination of a clear specular orientation, usually from the seismic image itself, misleading or ambiguous. The conversion of depth moveout into traveltime error along nonspecular rays was done according to an analytic mechanism, derived in the angle domain. The proposed analysis of migration errors by a multiple dip-angle orientation is demonstrated via the multidip tomography formulation by 2D synthetic and real data examples. It seems to be more efficient, as accurate and reliable as the conventional analysis, and to be better able to determine the ill-posed conditioning of the tomographic inversion.